Why is a check necessary when running cloud code without supervision?	To ensure the work done by the hands-off process is safe and reliable.
How should the level of verification be determined for a cloud run?	In proportion to the degree of unsupervision given to the run.
What is the danger of relying solely on a cloud summary of a change?	The summary might look fine while the underlying diff touches unexpected or problematic files.
What is the primary gate for validating an unsupervised run?	Whether the tests passed and whether the cloud system actually ran them, not just claimed they did.
What mechanisms should be used to enforce checks on unsupervised runs?	Hooks (like stop hooks or post tool use hooks) that feed failure back to the system.
What is the function of a hook in an automated process?	To fire on every run, ensuring checks happen even if the user forgets to ask.
How can you get a "cold second opinion" on an unsupervised change?	By running a subagent code review on the change with no memory of how it was built.
How should you verify the results of a headless run?	By checking the JSON result of the run and exiting the test.